It is thus important to map MR-accessible parameters that correlate well with the cortical myelin content (Hopf 1954, 1955, 1956) measured in cadaver brain sections. Among suggested candidates are proton density (Shah et al. 2011), myelin water fraction (MWF, Mackay et al. 2006), magnetization transfer rate (MTR, Wolff and Balaban 1989; Koenig 1991), transverse relaxation rate (R2*, Duyn 2011), phase maps formed using gradient echo imaging (Duyn et al. 2007), and longitudinal relaxation rate (R1) (Fischl et al. 2004). The challenge is to obtain a high enough spatial resolution to distinguish intra-cortical details that can assist in realistic parcellation, within a scanning time short enough for in-vivo studies of humans.

At this point it must be mentioned that currently there is no technique in existence that can quantitatively assay the density of myelin at high spatial resolution in cadaver brain tissue. The optical density of myelin stained sections of cadaver brain clearly reflects the amount of myelin, but this does not correspond to actual myelin density, because the microscope images are essentially shadow-grams, like flat X-rays, giving no information about the distribution of myelin in the third dimension perpendicular to the plane of the section. Chemical assays, such as those pioneered in the nineteenth century by Thudichum (1884), require volumes of tissue, at least 30 mm³, that preclude intracortical spatial resolution (Dasgupta and Hogan 2001).

Because the water content of grey and white brain tissue is about 80 % and 70 % respectively, it is possible to obtain useful contrast simply using maps of proton density (Shah et al. 2011), which demonstrate this very basic physical distinction. These can be obtained by means of very short echo time (TE) imaging, or by extrapolating sets of images with a range of echo times to zero TE. However, high signal to noise ratio (SNR) is required because of the relatively small grey-white difference in water content, and this method has not been widely used for distinguishing intra-cortical structures.

Myelin water fraction (MWF) is defined as the fraction of the NMR signal from water protons that can be unambiguously assigned to the thin layers of water that separate the layers of myelin membrane that wrap myelinated axons. There is good evidence (Kalantari et al. 2011) that this water exchanges very slowly with the other major pools of water in brain tissue, those within cell bodies and between cells. MWF can be estimated by determining the distribution of T₂ relaxation times (Whittall et al. 1997). In brain tissue, at 3 T magnetic field strength, T₂ distributions show three peaks, empirically assigned to compartmentalized spin populations: a short T₂ peak around 20 ms (between 10 and 50 ms) representing the water trapped between the layers of myelin, a second peak around 70–90 ms assigned to intra- and extracellular water, and a third peak with T₂ > 2 s usually assigned to cerebrospinal fluid signal. The myelin water fraction can be computed as the fraction of signal in the T₂ distribution below about 40 ms, typically representing about 15 % of the total water content. T₂ relaxation times can be obtained using the well-known Carr-Purcell-Meiboom-Gill technique (Kolind et al. 2009).

There are two practical limitations regarding the use of MWF for investigation of cortical myeloarchitecture. Firstly, as just mentioned, because there is no simple quantitative assay for myelin content in living human brain, it has not yet been demonstrated that MWF provides a linear measure of myelin density. The second limitation is the comparatively low signal-to-noise ratio (SNR) of accurate MWF methods, which means that either very large voxels or very long averaging times must be used. Although a faster sequence (mcDESPOT) based on steady-state free precession and fast gradient echo acquisition has been developed (Deoni et al. 2008), this method for quantifying MWF is very approximate, due to the large number of variables involved in the fitting procedure employed (Lankford and Does 2012), and the requirement of uniform transmit RF fields. For such reasons MWF mapping has not yet been used in cortical parcellation studies.

Measurement of magnetization transfer rate (MTR) has also been proposed as a method for myelin density quantification (e.g. Vavasour et al. 1998; Alexander et al. 2011). This method (Wolff 1989) relies on the fact that the spin–lattice relaxation of longitudinal magnetization in tissue arises largely from transfer of the spin magnetization of protons on free water molecules to protons bound to large relatively immobile biological molecules, mostly embedded in membranes. Transfer of spin magnetization in the opposite direction also occurs, from protons bound to the large molecules to the free water protons. This can be enhanced by applying an RF magnetic field at a frequency offset from the free water proton Larmor frequency, which excites the protons of hydrogen atoms attached to large molecules, having a broad resonance spectrum. The coupling of this spin magnetization to the free water protons accelerates their relaxation, and transfer of the magnetization saturates them, reducing the net magnetization. These effects on T1 and spin magnetization M₀ thus provide an indication of the local molecular environment. In much of brain tissue, especially white matter, myelin dominates the local molecular environment, and so it has been considered that MTR also provides a map of brain myelin (e.g. Gringel et al. 2009; Helms et al. 2010). However, while there is a rough concordance between maps generated using MWF and MTR methods, they do not match in detail (Vavasour et al. 1998), giving rise to a concern that neither of these methods attain this goal. MTR imaging also tends to provide a relatively low SNR per unit time, because it compares the signal from images with rather similar intensity, with and without a magnetization saturation pulse.

Given the ubiquity and concentration of myelin in brain tissue, it is expected to affect MR image contrast in many ways. Table 7.1 (Siegel et al. 1999) provides the dry weight composition for several important constituents of brain tissue.

It is clear from this table that myelin concentration levels are high enough to affect proton density contrast. Myelin also has a strong effect on the transverse relaxation times T2 and T2*. T2 is primarily affected because the water between the myelin layers wrapping axons has a particularly short T2, as described above. The T2 of the non-myelin water in white matter is surprisingly similar to T2 in grey matter (Whittall et al. 1997).

It is still poorly recognized how little T2 contrast there really is in the brain, outside of deep brain structures containing iron (see below). Because its volume fraction is small (less than 15 %), and its T2 is short ( < 15 ms), myelin water contributes little to the value of T2 measured conventionally by comparing the signal at two echo times, or by measuring with the Carr-Purcell sequence. Instead, so-called “T2-weighted images” provided by MRI sequences such as turbo-spin-echo (TSE) typically gain much of their grey-white matter contrast from magnetization transfer (see Thomas et al. 2004; Turner et al. 2008), differences in T1, or proton density variations. Optimal contrast-to-noise for TSE sequences, which typically have echo trains extending much longer than T2, is obtained at the shortest possible echo times, which would not be the case if T2 difference were the major contrast mechanism. When T1 was measured using an inversion-recovery prepared turbo-spin-echo sequence, it was found that MTC effects reduced its value by about 30 % (Turner et al. 2008). It has been historically unusual to create genuine maps of either the short or long T2 component of the bi-exponential transverse relaxation decay curve. A very careful study was performed by Oros-Peusquens et al. (2008). Her quantitative maps of T2 in normal brain show very little overall contrast between grey and white matter, with a large overlap in their T2 distributions. See also Fig. 7.2, which shows a quantitative map of T2 in normal brain obtained at the Leipzig MPI at 7 T. The futility of attempting cortical parcellation using T2 maps alone is underscored by West et al. (2012) who show the enormous overlap of the T2 distributions in WM and GM.

In practice, when magnetization transfer effects are carefully avoided, the grey-white contrast obtained by “T2-weighted sequences”, such as TSE, is dominated by variations in T1 and MTC at echo times longer than 100 ms. The main reason for this is that the echo train generated by the repeated radiofrequency refocussing pulses of the TSE sequence consists largely of stimulated echoes, for echo times longer than about 100 ms. Stimulated echoes decay with a time constant of T1, which is much longer than T2 in tissue, and thus the signal in these later echoes is heavily T1 weighted. Even with echo times as short as 26 ms (e.g. Trampel et al. 2011) TSE contrast is clearly dominated by magnetization transfer effects.

T2* weighted images and T2* maps of the brain, on the other hand, show highly significant signal variations, both between grey and white matter and within each of these tissues. This reflects the presence of myelin, but also another very important source of MRI contrast in brain, iron (Drayer et al. 1986). It is now becoming clear that the main origin of T2* contrast is that the magnetic susceptibility of lipid molecules, and of commonly found iron-containing molecules, particularly ferritin, is different from that of water. (Duyn 2011; Lee et al. 2011) When the tissue experiences a static magnetic field, this difference causes microscopic field gradients to arise in water associated with such molecules. This is especially the case in WM, with a high density of lipid membranes forming myelin sheaths around the axons, and in deep brain nuclei, such as the substantia nigra, with high densities of iron in the form of ferretin and neuromelanin. The net result of the dephasing of the spin magnetization caused by these magnetic field gradients is a reduction in T2*. This susceptibility difference also partly explains the strong phase contrast between GM and WM (Li et al. 2011). T2* weighted images and phase images of the brain can be obtained with very high spatial resolution. Cortex often contains a significant quantity of myelinated axons (the basis of myeloarchitecture, indeed), and iron deposits are often associated with such cortical myelin (Fukunaga et al. 2010) (see Fig. 7.3).

Thus T2*-weighted images show excellent detail in cortical grey matter (Duyn 2011) (see Fig. 7.2), which has been shown to depend on the angle between B₀ and the local axonal orientation (Cohen-Adad et al. 2012) (Fig. 7.4).

But probably the most striking impact of myelin on MR images lies in its effect on T1. For precisely the same reason that it affects MTR, the presence of myelin substantially accelerates T1 relaxation, by the transfer of magnetization from slowly relaxing free water protons to rapidly relaxing protons bound to large molecules in the cell membranes. Those specific large molecules that are particularly responsible for this powerful relaxation mechanism have been discovered by careful study of model membrane systems comprised of phospholipids and water (Koenig 1991; Kucharczyk et al. 1994). When either of two important components of typical cell membranes, cholesterol and cerebroside, were added in biologically typical concentrations, they were found to dramatically shorten T1. These molecules have hydrophilic hydrogen-containing head groups that sit at the membrane-water interface and provide a conduit for the transfer of magnetization from the free water protons to the bulk of the membrane.

Thus so-called “T1-weighted” images have come to be regarded as the standard for depicting grey-white contrast in human brain, and used in numerous applications which require image segmentation of grey and white matter, such as voxel-based morphometry (Ashburner and Friston 2000). Generally low flip-angle gradient-echo (FLASH) sequences (Frahm et al. 1986) are used to provide such images. Here the contrast arises because white matter protons recover faster than grey matter protons to alignment with the static field, providing a larger equilibrium signal. The contrast can be further enhanced using a preparatory inversion pulse, such that the grey matter signal is close to being nulled at the time of image data acquisition, at which time the white matter signal has already partially recovered. The resultant sequence (Mugler and Brookeman 1990) is named MP-RAGE (magnetization-prepared rapid-acquisition gradient echo). However, it should be pointed out that neither FLASH nor MP-RAGE sequences actually produce maps of the value of T1. They have higher intensity where T1 is shorter, and their intensity is also weighted by proton density and T2* values.

In order to obtain quantitative maps of T1 and thus to provide a better guide to tissue myelin, each of these sequences must be modified. Adding a second excitation pulse with a different flip angle to the FLASH sequence (Preibisch and Deichmann 2009), or sampling the MP-RAGE signal at two different inversion times, allows the longitudinal relaxation time T1 to be mapped with reasonable precision. The second of these sequences is known as MP2RAGE (Marques et al. 2010). Maps of T1 may also be generated by acquiring a series of inversion-recovery T1-weighted images, and fitting the inversion time dependence in each voxel to an exponential function. Importantly, these maps show little effect of the inhomogeneity of image intensity (bias-field) normally caused by spatial variations in RF excitation and reception.

In conclusion, comparing MRI methods for detecting the presence of myelin, it can be hypothesized that maps of relaxation rate R1 ( = 1/T1) and MTR maps provide essentially identical information, proportional to the area of membranes in good contact with the free water pool (i.e. the outer and inner circumferences of myelinated axons), while MWF maps provide a good quantification of total myelin density. This concept merits much deeper exploration, including quantitative chemical assays of myelin density, in future research.